# -------------------------------------------------------------------
# Purpose: Creates Table B31
# Author:  Max Posch, 25/07/2025
# Usage:   Source this script to generate the table.
# -------------------------------------------------------------------
# Check that required paths exist
stopifnot(dir.exists(pdataconfanalysis))
stopifnot(dir.exists(poutputappendix))


# Load data
load(file.path(pdataconfanalysis, "surnameCountyLevel19001940.RData"))


# Regressions
r <- list()
r <- append(r, list(feols(sum_patents_pc_1900_f_w ~ iv_lo_entropy_namelast_mp_adjp_ws | gisjoin_1900^namelast_mp + statefip^year, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
r <- append(r, list(feols(sum_patents_pc_1900_f_w ~ iv_lo_entropy_namelast_mp_adjp_ws + iv_lo_sum_n_namelast_mp_adjp_tr_ws | gisjoin_1900^namelast_mp + statefip^year, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
r <- append(r, list(feols(sum_patents_pc_1900_f_w ~ iv_lo_entropy_namelast_mp_adjp_ws + iv_lo_sum_n_namelast_mp_adjp_tr_ws | gisjoin_1900^namelast_mp + namelast_mp^year + statefip^year, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
r <- append(r, list(feols(sum_patents_pc_1900_f_w ~ iv_lo_entropy_namelast_mp_adjp_ws + iv_lo_sum_n_namelast_mp_adjp_tr_ws | gisjoin_1900^namelast_mp + namelast_mp^year + statefip^year + gisjoin_1900[year], surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
r <- append(r, list(feols(sum_break_p80_rrfsim05_pc_1900_f_w ~ iv_lo_entropy_namelast_mp_adjp_ws | gisjoin_1900^namelast_mp + statefip^year, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
r <- append(r, list(feols(sum_break_p80_rrfsim05_pc_1900_f_w ~ iv_lo_entropy_namelast_mp_adjp_ws + iv_lo_sum_n_namelast_mp_adjp_tr_ws | gisjoin_1900^namelast_mp + statefip^year, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
r <- append(r, list(feols(sum_break_p80_rrfsim05_pc_1900_f_w ~ iv_lo_entropy_namelast_mp_adjp_ws + iv_lo_sum_n_namelast_mp_adjp_tr_ws | gisjoin_1900^namelast_mp + namelast_mp^year + statefip^year, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
r <- append(r, list(feols(sum_break_p80_rrfsim05_pc_1900_f_w ~ iv_lo_entropy_namelast_mp_adjp_ws + iv_lo_sum_n_namelast_mp_adjp_tr_ws | gisjoin_1900^namelast_mp + namelast_mp^year + statefip^year + gisjoin_1900[year], surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))

i <- list()
i <- append(i, list(feols(sum_patents_pc_1900_f_w ~ 1 | gisjoin_1900^namelast_mp + statefip^year | entropy_namelast_mp_adjp_ws ~ iv_lo_entropy_namelast_mp_adjp_ws, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
i <- append(i, list(feols(sum_patents_pc_1900_f_w ~ 1 | gisjoin_1900^namelast_mp + statefip^year | entropy_namelast_mp_adjp_ws + sum_n_namelast_mp_adjp_ws ~ iv_lo_entropy_namelast_mp_adjp_ws + iv_lo_sum_n_namelast_mp_adjp_tr_ws, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
i <- append(i, list(feols(sum_patents_pc_1900_f_w ~ 1 | gisjoin_1900^namelast_mp + namelast_mp^year + statefip^year | entropy_namelast_mp_adjp_ws + sum_n_namelast_mp_adjp_ws ~ iv_lo_entropy_namelast_mp_adjp_ws + iv_lo_sum_n_namelast_mp_adjp_tr_ws, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
i <- append(i, list(feols(sum_patents_pc_1900_f_w ~ 1 | gisjoin_1900^namelast_mp + namelast_mp^year + statefip^year + gisjoin_1900[year] | entropy_namelast_mp_adjp_ws + sum_n_namelast_mp_adjp_ws ~ iv_lo_entropy_namelast_mp_adjp_ws + iv_lo_sum_n_namelast_mp_adjp_tr_ws, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
i <- append(i, list(feols(sum_break_p80_rrfsim05_pc_1900_f_w ~ 1 | gisjoin_1900^namelast_mp + statefip^year | entropy_namelast_mp_adjp_ws ~ iv_lo_entropy_namelast_mp_adjp_ws, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
i <- append(i, list(feols(sum_break_p80_rrfsim05_pc_1900_f_w ~ 1 | gisjoin_1900^namelast_mp + statefip^year | entropy_namelast_mp_adjp_ws + sum_n_namelast_mp_adjp_ws ~ iv_lo_entropy_namelast_mp_adjp_ws + iv_lo_sum_n_namelast_mp_adjp_tr_ws, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
i <- append(i, list(feols(sum_break_p80_rrfsim05_pc_1900_f_w ~ 1 | gisjoin_1900^namelast_mp + namelast_mp^year + statefip^year | entropy_namelast_mp_adjp_ws + sum_n_namelast_mp_adjp_ws ~ iv_lo_entropy_namelast_mp_adjp_ws + iv_lo_sum_n_namelast_mp_adjp_tr_ws, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))
i <- append(i, list(feols(sum_break_p80_rrfsim05_pc_1900_f_w ~ 1 | gisjoin_1900^namelast_mp + namelast_mp^year + statefip^year + gisjoin_1900[year] | entropy_namelast_mp_adjp_ws + sum_n_namelast_mp_adjp_ws ~ iv_lo_entropy_namelast_mp_adjp_ws + iv_lo_sum_n_namelast_mp_adjp_tr_ws, surnameCountyLevel19001940, weights = surnameCountyLevel19001940$wgt_name)))


# F-statistics
dstata <- surnameCountyLevel19001940[, .(
  x1 = entropy_namelast_mp_adjp_ws, 
  z1 = iv_lo_entropy_namelast_mp_adjp_ws,
  x2 = sum_n_namelast_mp_adjp_ws, 
  z2 = iv_lo_sum_n_namelast_mp_adjp_tr_ws,
  y1 = sum_patents_pc_1900_f_w, 
  y2 = sum_break_p80_rrfsim05_pc_1900_f_w,
  gisjoin_1900_f, statefip_f, year_f, year_num, namelast_mp_f, wgt_name
)]
commands <- list(
  'ivreghdfe y1 (x1= z1) [pw = wgt_name], absorb(gisjoin_1900_f#namelast_mp_f statefip_f#year_f) cluster(statefip_f) first ffirst savefirst
   gen swf1 = .
   replace swf1 = round(e(first)["SWF",1])
   keep swf*
   keep if _n == 1',
  'ivreghdfe y1 (x1 x2= z1 z2) [pw = wgt_name], absorb(gisjoin_1900_f#namelast_mp_f statefip_f#year_f) cluster(statefip_f) first ffirst savefirst
   gen swf1 = .
   gen swf2 = .
   replace swf1 = round(e(first)["SWF",1])
   replace swf2 = round(e(first)["SWF",2])
   keep swf*
   keep if _n == 1',
  'ivreghdfe y1 (x1 x2= z1 z2) [pw = wgt_name], absorb(gisjoin_1900_f#namelast_mp_f statefip_f#year_f year_f#namelast_mp_f) cluster(statefip_f) first ffirst savefirst
   gen swf1 = .
   gen swf2 = .
   replace swf1 = round(e(first)["SWF",1])
   replace swf2 = round(e(first)["SWF",2])
   keep swf*
   keep if _n == 1',
  'ivreghdfe y1 (x1 x2= z1 z2) [pw = wgt_name], absorb(gisjoin_1900_f#namelast_mp_f statefip_f#year_f year_f#namelast_mp_f c.year_num##gisjoin_1900_f) cluster(statefip_f) first ffirst savefirst
   gen swf1 = .
   gen swf2 = .
   replace swf1 = round(e(first)["SWF",1])
   replace swf2 = round(e(first)["SWF",2])
   keep swf*
   keep if _n == 1'
)
swf_results <- as.character(unlist(lapply(commands, get_fstat_from_stata, data.in = dstata)))
swfstat <- rep(swf_results, times = 2)


# Create table
x <- na.omit(surnameCountyLevel19001940[, .(sum_patents_pc_1900_f_w, wgt_name)])
y1_mean <- round(weighted.mean(x$sum_patents_pc_1900_f_w, x$wgt_name), 2)
y1_sd <- round(sqrt(weighted.mean((x$sum_patents_pc_1900_f_w - y1_mean)^2, x$wgt_name)), 2)
x <- na.omit(surnameCountyLevel19001940[, .(sum_break_p80_rrfsim05_pc_1900_f_w, wgt_name)])
y2_mean <- round(weighted.mean(x$sum_break_p80_rrfsim05_pc_1900_f_w, x$wgt_name), 2)
y2_sd <- round(sqrt(weighted.mean((x$sum_break_p80_rrfsim05_pc_1900_f_w - y2_mean)^2, x$wgt_name)), 2)
y1 <- paste0("\\makecell{Patents \\\\ per 1,000 people \\\\ (mean = ", y1_mean, ", sd = ", y1_sd, ")}")
y2 <- paste0("\\makecell{Breakthrough patents \\\\ per 1,000 people \\\\ (mean = ", y2_mean, ", sd = ", y2_sd, ")}")

setFixest_dict(
  c(
    entropy_namelast_mp_adjp_ws = "Surname diversity",
    iv_lo_entropy_namelast_mp_adjp_ws = "Predicted surname diversity",
    sum_n_namelast_mp_adjp_ws = "Population",
    iv_lo_sum_n_namelast_mp_adjp_tr_ws = "Predicted population",
    sum_patents_pc_1900_f_w = y1, sum_break_p80_rrfsim05_pc_1900_f_w = y2,
    year = "Period", statefip = "State", namelast_mp = "Surname", gisjoin_1900 = "County"
  )
)

tablename <- file.path(poutputappendix, "tableB31.tex")
etable(r,
  cluster = ~statefip,
  fitstat = ~n,
  digits = "r3", digits.stats = "r3",
  file = tablename, replace = TRUE,
  style.tex = style.tex("aer"), tex = TRUE
)
edit_table_content_fixed(tablename, "Period $\\times $ County", "County-specific linear trends")
add_table_row(tablename, "\\midrule", "\\multicolumn{2}{l}{\\textit{Panel A: Reduced-form estimates}} &  \\multicolumn{7}{c}{}\\\\ \\cmidrule(lr){1-9}")
add_table_row(tablename, "mean =", "\\cmidrule(lr){2-5}  \\cmidrule(lr){6-9}")
move_table_row(tablename, "Observations", "bottomrule")
add_table_row(tablename, "    \\\\", c("\\multicolumn{2}{l}{\\textit{Panel B: Instrumental-variable estimates}} &  \\multicolumn{7}{c}{}\\\\", "\\\\", "\\multicolumn{2}{l}{\\textit{Panel C: First-stage estimates}} &  \\multicolumn{7}{c}{}\\\\"))

temptable <- file.path(poutputappendix, "temp.tex")
etable(i,
  cluster = ~statefip,
  fitstat = ~n,
  digits = "r3", digits.stats = "r3",
  extralines = list("Sanderson-Windmeijer \\textit{F}-stat" = swfstat),
  file = temptable, replace = TRUE,
  style.tex = style.tex("aer"), tex = TRUE
)
estimates_rows <- get_estimates_rows(temptable)
add_table_row(tablename, "Panel B", c("\\cmidrule(lr){1-9}", estimates_rows))
add_table_row(tablename, "Sanderson", "\\\\", "before")

temptable <- file.path(poutputappendix, "temp.tex")
etable(i,
  stage = 1,
  cluster = ~statefip,
  fitstat = ~n, digits = "r3", digits.stats = "r3",
  file = temptable, replace = TRUE,
  style.tex = style.tex("aer"), tex = TRUE
)
estimates_rows <- get_estimates_rows(temptable)
estimates_rows <- collapse_stage1(estimates_rows, c(2, 3, 5, 7, NA, 4, 6, 8))
add_table_row(tablename, "Panel C", estimates_rows)
add_table_row(tablename, "Panel C", c("\\cmidrule(lr){1-9}", "& \\multicolumn{4}{c}{Surname diversity} & \\multicolumn{4}{c}{Population}\\\\", "\\cmidrule(lr){2-5}  \\cmidrule(lr){6-9}"))
remove_table_row(tablename, "County fixed effects")
file.remove(temptable)

cat("Table B31 saved to:", tablename, "\n")